Sustainable material design using universal-AI

December 23, 2025

New technology that comes close to resolving the trade-off between degradability and toughness

Researchers from the RIKEN CSRS conducted a systematic analysis of biodegradable plastics using a self-developed multimodal–multitask deep learning model and revealed a potential relationship between the degradation process by microorganisms and the toughness of polymer materials.

Biodegradable plastics face a trade-off: they need sufficient toughness when used, while they are preferred to degrade when unexpectedly drained into the coastal environment. To resolve this issue, the research team focused on the relationships between the materials and their degradability and mechanical properties at the molecular level. They extracted the chemical structures and molecular motion of biodegradable plastics as thoroughly as in health checkups, using various nuclear magnetic resonance techniques. They extracted essential factors that could influence on degradability and mechanical properties by integrating universal-AI modeling. By integrating the molecular structures, thermal properties, and multiple dynamic data obtained through NMR techniques, they revealed multiple characteristics such as flexibility, segment motility, and partial molecular structures of molecular chains are primary factors to define both degradative rates and toughness. These findings suggest that these molecular-level structures and dynamic properties are the keys to determining the trade-off between degradability and toughness.

From now on, leveraging this machine learning model, or artificial general intelligence, could not only reveal precise relationships between the structures and physical properties of valuable biodegradable plastics but also identify hierarchical characteristics that impact the balance of material performances. Further efforts will help develop practical guidelines for rationally designing sustainable materials.

 

Original article

Sustainable Materials and Technologies　doi: 10.1016/j.susmat.2025.e01781

X. Ni、Y. Amamoto、J. Kikuchi,

"Simultaneous multimodal and multitask strategies for diverse biodegradable polymers powered by NMR data science".

Contact

Jun Kikuchi; Team Director
Xinyu Ni; Junior Research Associate
Yoshifumi Amamoto; Visiting Scientist
Environmental Metabolic Analysis Research Team